Diffusion tensor magnetic resonance (MR) imaging of the brain combined with machine learning could provide a fast and objective alternative to traditional methods of diagnosing autism spectrum disorder (ASD), according to a recently published study.
Currently, the gold standard for an autism diagnosis is a set of diagnostic evaluations, such as the Autism Diagnostic Observation Schedule or Autism Diagnostic Interview-Revised report. A diagnosis requires repeated visits to a pediatric specialist. Because the diagnosis is subjective, the score can vary from one specialist to another.
In a recently published study in the journal Sensors, researchers were motivated to develop a neuroimaging-based alternative that can provide an objective evaluation, which could help clinicians reach a diagnosis faster and more reliably.
“The results are promising. This is an important step toward using the latest technology in a scientifically rigorous way to diagnose autism,” said Gregory N. Barnes, M.D., Ph.D., co-author of the study, “The Role of Diffusion Tensor MR Imaging (DTI) of the Brain in Diagnosing Autism Spectrum Disorder: Promising Results.” Dr. Barnes is a pediatric neurologist with Norton Children’s Autism Center and Norton Children’s Neuroscience Institute, both affiliated with the UofL School of Medicine.
Dr. Barnes’ co-authors included researchers from the Bioengineering Department at the University of Louisville and the Department of Electrical and Computer Engineering at Abu Dhabi University in the United Arab Emirates.
ASD affects approximately 2% of U.S. children, but the pathology of the disorder remains unknown. Differences in brain development, and environmental and/or genetic factors are thought to play a role.
Refer a patient
To refer a patient to the Norton Children’s Autism Center, visit Norton EpicLink and choose EpicLink referral to Pediatric Autism.
Mapping the brain’s white matter connecting cortical and subcortical regions
Symptoms of ASD can vary considerably. Common symptoms are difficulty with social interactions and communications, poor eye contact or avoidance of eye contact, repetitive or ritualistic behaviors, and self-injury.
Previous neuroimaging studies looking at autism have focused on anomalies in brain shape or volume using structural MRI, anomalies in white matter diffusivity using diffusion tensor MR imaging, or task- or function-based functionality using fMRI.
Diffusion tensor MR imaging uses the rate of water diffusion between cells to create a 3D map to generate more detailed images of the brain than a conventional MRI.
Researchers in this study developed a diffusion tensor imaging-based algorithm using machine learning to look for insights from brain imaging of white matter connectivity. Diffusion tensor MR imaging enables 3D modeling of the white matter organization in the brain. White matter mainly consists of axons of neurons, with these axonal fibers carrying signals between various brain regions and between the brain and spinal cord.
Dr. Barnes and his colleagues verified their proposed framework using a large, publicly available diffusion tensor imaging dataset of 225 individuals from the Autism Brain Imaging Data Exchange-II (ABIDE-II). The data comprised 125 autistic subjects and 100 typically developed ones.
Looking at the data, the researchers identified 12 white matter brain area pairs that were important and aligned with previous literature studying autism impairments. All the white matter tracts identified in the study connect cortical and subcortical regions, contributing to the inattention, self-injury, repetitive behaviors, motor, social, memory, emotional regulation and sensory impairments found in autistic individuals.
“Adding more datasets should guarantee generalizability of our proposed framework, which can be a good direction for future work,” the co-authors concluded. Combining diffusion tensor MR imaging with other tools such as fMRI or sMRI also may help, according to the authors.